Data analytics in IIoT: Edge vs. cloud

Data analytics in IIoT: Edge vs. cloud
It goes without saying that plants and factories have become smarter due to the Industrial Internet of Things. At the center of IIoT is data, which helps operators gain better insights into the manufacturing facility. While the analysis of data is commonly done on a server in the backend, more and more this is done on the edge.
In IIoT, data is king. Piles upon piles of data generate by IIoT devices provide visibility for the factory operator. “Traditionally, and for decades, sensor data has been mainly measured for three purposes: as inputs to the real-time control process of the machine/factory; to display it in control rooms, so human controllers can view and react to various changes, or be alerted by simple rules; and stored locally for post-mortem purposes and used after a failure for analysis and re-adjustment of control processes,” said Eitan Vesely, CEO of Presenso.
Specifically, data can help with preventive maintenance, which has become critical for manufacturers in their efforts to minimize factory downtime. “We believe that hidden within terabytes of stored machine-generated sensor data are micro-patterns that can warn us of machine failure. Until a couple of years ago, they were undetectable by even the most advanced monitoring systems. Data scientist lacked the tools to predict machine failure and industrial plants lacked the data scientists to even try. With Industry 4.0 we are seeing a surge of interest in IIoT predictive maintenance solutions and a number of solution providers enter the market,” Vesely said.

Edge vs. cloud

Once the data is generated and collected, analyzing it becomes key. In the more conventional approach, data is sent to the server in the backend/cloud for analysis. But more and more, this process has been moved to the edge, which holds various advantages.
“Time-sensitive data analysis needed in the millisecond to the sub-second area is better processed at the edge because if there is an issue, corrective actions must be addressed immediately and should have minimal latency,” said Eric Ehlers, Vertical Marketing Manager at Cisco. “Edge networking helps better streamline bandwidth and latency by enabling decision making close to the point of action. It also optimizes the sending of data on the network. Policies can then be created that ensure the right data gets to the right people at the right time, thus providing meaningful information and insights.”
“Analytics on edge devices is required when decisions need to be made that are time critical, where the results of analytics need to be applied immediately in the manufacturing process to correct for problems detected during manufacture,” said Howard Wang, Director of Sales for APAC/ROW at Real-Time Innovations (RTI). “Edge analytics are important because the faster one can process information from sensors in the manufacturing line and make adjustments/decisions, the sooner issues that affect quality and production can be dealt with, saving money, increasing output, improving quality, and optimizing the production process.”
But this is not to say that analytics on the server in the backend is no longer needed. “Analytics in the cloud is needed for long-term analyses which are not time critical.  Cloud analytics is usually better for calculations that take significant time to complete and operate on large quantities of data – that is, big data. Example uses are predictive maintenance and process optimization,” Wang said.
“We stream data from all the sensors within a production environment to our server in real time. Edge may be more suitable for hardware-specific point solutions, but if data from an entire facility needs to be processed and analyzed, then cloud is most effective,” Vesely said.
In fact, the best way to optimize IIoT in a manufacturing environment is to combine both approaches. "For comprehensive analysis of data, a combination of edge analytics along with in-premise servers is befitting in the current scenario. Edge devices enabled with artificial intelligence (AI) and machine learning algorithms can capture and analyze critical data. This activity allows in sending only the processed data to the servers. It also makes further analysis of data on the cloud/premise servers more efficient, reducing the data latency and enabling the sensor data to seamlessly work with control system. The combination of edge analytics along with in-premise servers reduces bandwidth clogging as only the processed data is pushed into the servers. Also, the alarms and events captured by edge devices are available to workers and maintenance staff operating on the machines," said Keshab Panda, CEO and Managing Director of L&T Technology Services.
“Mission-critical applications that need low latency and need to be measured quickly are better suited at the edge. Applications with large volumes of data that will require more computing power are better suited in server/cloud environments. When you combine real-time data analysis at the edge and big data analytics of historical data then you can identify efficiencies that can help improve quality, yield and overall equipment effectiveness (OEE). This data can also be integrated into enterprise resource planning (ERP) systems and manufacturing execution systems (MES) to support better utilization, visibility and help ensure more quality control across production runs,” Ehlers said.

Product Adopted:
Share to:
Comments ( 0 )